Sentimony Skills - Setup Guide¶
Source: sentimony/skills via skills.sh - 28.3K combined installs across 25 indexed listings; first seen Aug 18, 2026 (evening sweep); converted Oct 10, 2026 (zero-catalog audit)
GitHub: sentimony/skills (6 stars, MIT license; pushed Oct 10, 2026; plugins/<plugin>/skills/<name>/SKILL.md layout)
Category: Dev Workflow / Agent Discipline
Quality Tier: 🟡 Beta - 25-skill dev-workflow suite; plugin layout; mixed sampled verdicts - see Security
Sentimony's skills repo packages developer discipline for coding agents into five plugins: devflow (14 skills), skills (9), writing (3), echarts, and skill-crafting. It has drawn 28.3K combined installs, and its busiest entries are the daily utilities: typescript (2,595), web-debug (2,361), vitest (2,347), and echarts (2,294). Underneath sits an opinionated loop: scope-triage classifies a request before design work, plan-crafting turns settled requirements into a TDD-oriented plan, inline-plan-dev or subagent-plan-dev executes it, review-request and review-resolution carry the review loop, and branch-finish closes verified work.
Skills live at plugins/<plugin>/skills/<name>/SKILL.md, and they install either individually through the skills CLI or as Claude Code and Codex plugins. Plugin skills are namespaced by their plugin (/devflow:scope-triage), and the workflow is composable rather than mandatory-linear: steps like debugging, git-worktree-isolation, and parallel-agents are reached when the work calls for them.
Installation¶
Prerequisites: Node.js for the skills CLI, or Claude Code / Codex for the plugin route.
# Install everything for Codex and Claude Code
npx skills add sentimony/skills -a codex claude-code -y
# Install a single skill
npx skills add sentimony/skills -s typescript -a codex claude-code -y
Or install as plugins, which namespaces every skill:
# Claude Code
claude plugin marketplace add sentimony/skills
claude plugin install devflow@sentimony
# Codex
codex plugin marketplace add sentimony/skills
codex plugin add devflow@sentimony
Other plugins in the set: writing, skill-crafting, echarts, and skills.
What It Provides¶
Install counts from the Oct 10, 2026 skills.sh snapshot:
| Skill | Installs | Use For |
|---|---|---|
| typescript | 2,595 | Configure tsconfig, diagnose compiler behavior, and audit or migrate TypeScript projects |
| web-debug | 2,361 | Debug and verify local web apps via Playwright |
| vitest | 2,347 | Configure, write, debug, run, migrate, and audit Vitest tests |
| echarts | 2,294 | Build, audit, style, debug, and optimize Apache ECharts visualizations |
| plan-crafting | 2,051 | Turn an approved design into a bite-sized, TDD-oriented implementation plan |
| scope-triage | 2,050 | Classify request scope before design work and route to the right process |
| dashfix | 1,919 | Replace typographic dashes with the plain hyphen and audit a project per occurrence |
| negafix | 1,910 | Ban negative parallelism, audit prose for it, and score it 0-100 |
| commit-all | 1,671 | Gather the working tree into one commit per repository the session changed |
| maintaining-agent-context | 1,527 | Audit and maintain a repo's agent instruction architecture for Claude Code and Codex |
| frontend-crafting | 1,200 | Create, redesign, review, and polish interfaces with a verifiable quality gate |
| review-request | 627 | Prepare and dispatch independent code review against requirements, scope, and the diff |
| tdd | 625 | Drive behavior changes through valid RED, sufficient GREEN, and evidence-backed refactoring |
| parallel-agents | 625 | Prove work units independent and dispatch one bounded parallel wave |
| subagent-plan-dev | 623 | Execute a plan through scoped subagents with risk-based dispatch and verification |
| review-resolution | 623 | Validate and resolve review findings with evidence and explicit dispositions |
| inline-plan-dev | 623 | Execute a plan inline in the current session with reconciliation and durable resume |
| git-worktree-isolation | 623 | Select, detect, or create a safe isolated workspace with explicit ownership |
| debugging | 623 | Investigate bugs with a root-cause-first evidence workflow |
| branch-finish | 622 | Decide, execute, and report the integration outcome for verified work |
The remaining 5 indexed listings range from 85 to 330 installs, led by prose-crafting (330) and secret-hygiene (132).
Why This Matters for Hermes Agents¶
Most agent failures are process failures: work that was never scoped, plans that drifted from reality, reviews that rubber-stamped a diff, branches merged without a verification pass. This suite encodes the countermeasures as skills a Hermes agent can reach for at each stage, from scope-triage at the front to branch-finish at the end, with review loops that demand evidence instead of assertions. The utilities stand alone as well: typescript, vitest, and web-debug cover the daily TypeScript and browser work, while dashfix and negafix police prose quality. Because everything ships per-skill and per-plugin, adoption can be gradual: start with scope-triage and commit-all, then add the review loop when the team is ready. The plugin namespacing pattern is also a clean model for Hermes skill authors building their own multi-plugin distributions.
Usage¶
| You say | What happens |
|---|---|
| "Is this change small enough to just do?" | scope-triage classifies scope and routes to direct work, a light spec, or a full design cycle |
| "Turn this spec into an implementation plan" | plan-crafting produces a bite-sized, TDD-oriented plan |
| "Execute the plan in this session" | inline-plan-dev works through the plan with plan-reality reconciliation and durable resume |
| "Run this plan through subagents" | subagent-plan-dev dispatches scoped subagents with risk-based verification |
| "Get an independent review of this diff" | review-request prepares and dispatches review against requirements and the actual diff |
| "Debug why the settings page fails to load" | web-debug drives Playwright to reproduce the issue and gather browser evidence |
| "Commit everything I changed today" | commit-all gathers the working tree into one commit per changed repository |
Verification¶
# Confirm the install through the skills CLI
npx skills list | grep -i sentimony
# Or check the plugin layout in a checkout of the repo
ls plugins/devflow/skills/
# Review the typescript skill from GitHub before installing
curl -sL https://raw.githubusercontent.com/sentimony/skills/main/plugins/skills/skills/typescript/SKILL.md | head -20
Security¶
skills.sh verdicts for sampled skills (verified Oct 10, 2026) - re-check the security pages on skills.sh before production use:
| Skill | Gen Agent Trust Hub | Socket | Snyk |
|---|---|---|---|
| typescript | Pass | Pass | Pass |
| web-debug | Pass | Pass | Warn |
| vitest | Pass | Pass | Pass |
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations¶
- Small repository: 6 GitHub stars and an independent publisher; the workflow suite is opinionated and young.
- Mixed sampled verdicts: web-debug carries a Snyk Warn, and coverage is partial across the 25 listings.
- Plugin migration caveat: from v1.56.0 the skills@sentimony plugin carries nine skills; the other 19 moved to devflow, writing, skill-crafting, and echarts, so install the plugins whose skills you use.
- The README lists 28 skills across five plugins while this snapshot indexes 25 listings; the experimental scope-check and webapp-debugger sit outside the index, so verify names when scripting installs.
- Snapshot data, verified Oct 10, 2026: 28,260 combined installs across 25 indexed listings; 6 GitHub stars; MIT; last pushed Oct 10, 2026. Counts drift over time.
Related¶
- Meticulous Agent Skills - Visual Regression Testing Setup - visual regression checks to pair with frontend-crafting and web-debug
- Skill Vetter - Security Audit for Hermes Skills Setup - vet any skill's security before you install it
- TanStack Skills - React Server State & Router Suite Setup - frontend stack skills that complement the typescript and vitest workflow
- Skills Catalog | Skills Marketplace
Pro Tips¶
- Start with scope-triage: it is the documented entry point, and every step after it is reached on a condition rather than in sequence.
- Prefer plugin installs in Claude Code or Codex to get namespaced commands like
/devflow:scope-triage. - Pair plan-crafting with inline-plan-dev for solo work; switch to subagent-plan-dev when the work parallelizes.
- dashfix and negafix are cheap prose gates; run them over docs and PR descriptions before review.
- Install single utilities with
-s(for example-s vitest) when you do not want the full workflow surface.
← Skills Catalog | Skills Marketplace → Powered by CorpusIQ